ft_0126_kor_eng-1

This model is a fine-tuned version of facebook/wav2vec2-xls-r-300m on the None dataset. It achieves the following results on the evaluation set:

• Loss: 1.0609
• Cer: 0.2143

Model description

More information needed

Intended uses & limitations

More information needed

Training and evaluation data

More information needed

Training procedure

Training hyperparameters

The following hyperparameters were used during training:

• learning_rate: 0.0001
• train_batch_size: 8
• eval_batch_size: 8
• seed: 42
• optimizer: Adam with betas=(0.9,0.999) and epsilon=1e-08
• lr_scheduler_type: linear
• lr_scheduler_warmup_steps: 100
• num_epochs: 40
• mixed_precision_training: Native AMP

Training results

Training Loss	Epoch	Step	Validation Loss	Cer
9.5689	2.44	100	6.7682	1.0
3.0585	4.88	200	4.9715	1.0
2.9664	7.32	300	4.7381	1.0
2.9563	9.76	400	4.3435	1.0
2.9441	12.2	500	3.8964	1.0
2.927	14.63	600	3.7284	1.0
2.9019	17.07	700	3.4735	1.0
2.7333	19.51	800	2.8455	0.9113
2.1073	21.95	900	1.8594	0.5217
1.488	24.39	1000	1.4387	0.3741
1.1468	26.83	1100	1.1822	0.2872
0.9639	29.27	1200	1.1472	0.2511
0.8441	31.71	1300	1.0706	0.2275
0.808	34.15	1400	1.0858	0.2239
0.6775	36.59	1500	1.0570	0.2136
0.6496	39.02	1600	1.0609	0.2143

Framework versions

• Transformers 4.35.2
• Pytorch 2.1.1+cu121
• Datasets 2.13.0
• Tokenizers 0.15.0